Mileage positioning system of railway inspection vehicle

Abstract

The invention discloses a mileage positioning system of a railway inspection vehicle, and relates to the technical field of unmanned driving. The system comprises a sleeper image snapshot module and a data analysis module, wherein the sleeper image snapshot module is used for shooting a sleeper image once every preset driving distance; and the data analysis module is used for counting the total number of sleepers in a target driving period according to the sleeper images, and determining mileage information in the target driving period according to the total number of sleepers. According to the technical scheme provided by the invention, more accurate mileage positioning is realized, and the problem of positioning loss is not easy to occur.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of driverless technology, and in particular to an odometer positioning system for a railway inspection vehicle. Background Art

[0002] Currently, for railway flaw detection and maintenance tasks, there are both large and high-speed dedicated railway maintenance track vehicles, as well as small and medium-sized low-speed track inspection vehicles or instruments. The former has a high speed and a long detection distance, while the latter has a small inspection range and is more delicate. The small and medium-sized low-speed track inspection vehicle integrated with autonomous driving is a new product in recent years. Currently, there are still many problems in the odometer link of the product.

[0003] Common odometer positioning methods mainly include: 3D point cloud positioning by lidar, GPS positioning, and positioning by integrating the motor speed or encoder speed, etc. Since the ground railway along the line is very open, the 3D point cloud positioning method by lidar is prone to losing positioning due to similar environments. In ground bridge holes, underground tunnels and other places, GPS signals are easily lost, and in such an environment, the GPS positioning method cannot calibrate and restore the accurate position. And the speed integration method itself has a large cumulative error, cannot be corrected, and only has distance information without position information. Summary of the Invention

[0004] The embodiments of the present invention provide an odometer positioning system for a railway inspection vehicle to solve the problems of easy loss of positioning and large positioning errors in the prior art.

[0005] The embodiments of the present invention provide an odometer positioning system for a railway inspection vehicle, and the system includes: a sleeper image capture module and a data analysis module; wherein,

[0006] The sleeper image capture module is used to capture a sleeper image every time a preset driving distance is traveled;

[0007] The data analysis module is used to count the total number of sleepers during the target driving period according to the sleeper image, and determine the mileage information during the target driving period according to the total number of sleepers.

[0008] Optionally, the sleeper image capture module includes an encoder, a microcontroller, and a camera; wherein,

[0009] The encoder is used to generate a pulse signal based on the rotation of the wheels of the inspection vehicle;

[0010] The microcontroller is used to count the driving distance according to the pulse signal, and generate a shooting trigger signal every time the preset driving distance is counted;

[0011] The camera is used to trigger the capture of the sleeper image according to the shooting trigger signal and upload it to the data analysis module.

[0012] Optionally, the data analysis module is specifically configured to:

[0013] Determine the mileage information according to the product of the total number of sleepers and the standard sleeper spacing.

[0014] Optionally, the two actual railway areas corresponding to two adjacent sleeper images are continuous; the data analysis module is specifically configured to:

[0015] Use an artificial intelligence detection and recognition algorithm to real-time recognize the number of sleepers in each sleeper image;

[0016] Sum up the number of sleepers, and if sleepers are respectively recognized at the joints of two adjacent sleeper images, subtract one from the statistical quantity to obtain the total number of sleepers.

[0017] Optionally, the system further includes a GPS positioning module for mileage positioning.

[0018] Optionally, the target driving period is the period during which the GPS positioning module has abnormal positioning.

[0019] Optionally, the GPS positioning module adopts a positioning method that combines GPS and high-precision inertial navigation.

[0020] Optionally, the system further includes a railway inspection module for detecting railway defects.

[0021] Optionally, the data analysis module is further configured to receive the defect information sent by the railway inspection module, and record the corresponding defect information together with each preset number of sleeper images collected before and after the corresponding moment.

[0022] Optionally, the data analysis module is further configured to detect the fasteners and sleeper defects according to the sleeper images.

[0023] An embodiment of the present invention provides a mileage positioning system for a railway inspection vehicle. The system includes a sleeper image capture module and a data analysis module. The sleeper image capture module captures a sleeper image every preset driving distance, and the data analysis module counts the total number of sleepers during the target driving period according to the sleeper images, and then determines the mileage information during the target driving period according to the total number of sleepers. More accurate mileage positioning can be achieved, and the problem of positioning loss is not likely to occur. Description of the Drawings

[0024] Figure 1 It is a schematic structural diagram of the mileage positioning system for a railway inspection vehicle provided by an embodiment of the present invention;

[0025] Figure 2 It is an exemplary system topology diagram provided by an embodiment of the present invention;

[0026] Figure 3 It is a schematic diagram of the neural network structure adopted in the embodiments of the present invention. Detailed implementation manners

[0027] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the present invention, rather than limiting the present invention. In addition, it should be noted that for the convenience of description, only parts related to the present invention rather than all structures are shown in the drawings.

[0028] Figure 1 It is a schematic diagram of the mileage positioning system of the railway inspection vehicle provided by the embodiments of the present invention. This embodiment is applicable to the situation of assisting an autonomous driving inspection vehicle in mileage positioning. As Figure 1 shown, the system includes: a sleeper image capture module 100 and a data analysis module 200; wherein, the sleeper image capture module 100 is used to capture sleeper images at every preset driving distance; the data analysis module 200 is used to count the total number of sleepers during the target driving period according to the sleeper images, and determine the mileage information during the target driving period according to the total number of sleepers.

[0029] Specifically, the number of sleepers itself can provide certain positioning basis for the staff, that is, the total number of sleepers obtained by statistics can be used as the mileage information during the target driving period. At the same time, sleepers usually have a certain standard sleeper spacing, so the driving mileage can also be calculated through the total number of sleepers obtained by statistics as the mileage information during the target driving period. Among them, the target driving period can be the entire inspection process, or a part of the journey that needs to be positioned in this way during the inspection. The sleeper image capture module 100 can keep capturing sleeper images throughout the inspection process, or only capture sleeper images within this part of the journey.

[0030] During the driving process of the inspection, the total number of sleepers during the target driving period can be obtained by continuously capturing sleeper images and performing sleeper recognition. Specifically, the sleeper image capture module 100 can capture a sleeper image at every preset driving distance, and when each sleeper image is captured, the sleeper image can be uploaded to the data analysis module 200 for storage. The data analysis module 200 can, when receiving each sleeper image, perform real-time recognition on the sleeper image to obtain the number of sleepers therein, and can perform real-time accumulation. At the same time, during the accumulation process, the counting of repeatedly appearing sleepers can be excluded, such as not counting the sleepers that have appeared in the current sleeper image in history. Among them, the preset driving distance and the sleeper image capture area can be adaptively set so that each passing sleeper can be captured, thereby ensuring the complete statistics of the total number of sleepers during the target driving period.

[0031] Based on the above technical solution, optionally, the sleeper image capture module 100 includes an encoder, a microcontroller, and a camera; wherein, the encoder is configured to generate a pulse signal based on the rotation of the wheels of the inspection vehicle; the microcontroller is configured to count the traveling distance according to the pulse signal, and generate a shooting trigger signal once every time the preset traveling distance is counted; the camera is configured to trigger the shooting of the sleeper image according to the shooting trigger signal and upload it to the data analysis module 200.

[0032] Specifically, as Figure 2 shown, the encoder can be set on the wheel rotation axis of the inspection vehicle, and can generate a corresponding pulse signal based on the wheel rotation and provide it to the microcontroller. The microcontroller can select stm32 (such as STM32F407), and can count the rising edges of the pulse signal through the built-in counter. According to the wheel diameter and the encoder resolution, it can be determined how many pulse signal rising edges need to be counted when the vehicle travels to the preset traveling distance. Exemplarily, the wheel diameter (the distance from the rotation axis to the track contact surface) is 48.8 cm, and the encoder resolution is 5000 P / R. Then the actual distance for the encoder to rotate one grid is 48.8 * 3.14 / 5000 = 0.3067 (mm). The preset traveling distance can be set to 50 cm, so 50 cm / 0.3067 mm = 1630 rising edges are required. When the microcontroller counts the corresponding number of rising edges, it can be considered that the inspection vehicle has traveled to a preset traveling distance once, that is, a shooting trigger signal is generated and sent to the camera, thereby triggering a shooting. Subsequently, the count can also be cleared for the next count. The camera can select a high-precision CMOS area array camera, and can be set at the bottom of the inspection vehicle facing directly downward. Each time it receives a shooting trigger signal, it can take a shot and upload the captured sleeper image to the data analysis module 200. The data analysis module 200 can select an industrial computer.

[0033] Among them, taking the microcontroller selecting stm32 as an example, its internal parameters can be set as follows: the A phase of the encoder is connected to the PE9 pin, the B phase of the encoder is connected to the PE11 pin, both pins are set to pull-down resistors, AF (reuse) mode. Within one jump cycle of the encoder, the timer of STM32 will count 4 times. The STM32 internally sets the output voltage value of the PC1 pin to be inverted once after the timer interrupt overflows. Therefore, two overflow interrupts are required to generate a rising edge. So the ARR parameter value of timer 1 is set to 1630 * 4 / 2 = 3260, the PSC parameter is set to 0, the encoder interface of the timer is set to encoder mode TI12 (both A and B phases count), rising edge counting, the filter parameter is set to 10, and the GPIO of the low-frequency output signal is configured as the PC1 pin, and the pull-up resistor push-pull output mode is set.

[0034] Based on the above technical solution, optionally, the data analysis module 200 is specifically configured to: determine the mileage information according to the product of the total number of sleepers and the standard sleeper spacing. Specifically, the actual driving mileage during the target driving period can be estimated by the product of the total number of sleepers and the standard sleeper spacing counted during the target driving period, and the estimated driving mileage can be used as the mileage information. The standard sleeper spacing can be determined by the actual railway design and the inspection section. Exemplarily, the standard sleeper spacing is 60 cm. Further, due to possible sleeper installation errors and the influence of factors such as curved tracks and switches, there is a certain error between the estimated driving mileage and the actual driving mileage. Then, an estimated mileage range can also be determined according to the estimated driving mileage as the mileage information.

[0035] Based on the above technical solution, optionally, the two actual railway areas corresponding to two adjacent sleeper images are continuous; the data analysis module 200 is specifically configured to: use an artificial intelligence detection and recognition algorithm to real-time recognize the number of sleepers in each sleeper image; sum up the number of sleepers, and if sleepers are respectively recognized at the connection of two adjacent sleeper images, subtract one from the statistical quantity to obtain the total number of sleepers. Specifically, the sleeper images uploaded by the sleeper image capture module 100 can be used as the input of the artificial intelligence detection and recognition algorithm. Through the artificial intelligence detection and recognition algorithm, the sleepers existing in the sleeper image can be recognized, and the number of sleepers therein can be counted. For example, the positions of the sleepers in the sleeper image can be marked by rectangular frames, and the final number of rectangular frames is the number of sleepers in the corresponding sleeper image. The artificial intelligence detection and recognition algorithm can be pre-trained with a large number of positive sample images with sleepers and corresponding annotations and negative sample images without sleepers. Specifically, when entering the target driving period, the counting of the number of sleepers can be started, and the number of sleepers in the sleeper images real-time uploaded by the sleeper image capture module 100 is real-time recognized and accumulated. The positions and parameters of the cameras in the sleeper image capture module 100 and the setting of the preset driving distance can be used to make the two actual railway areas corresponding to two adjacent sleeper images continuous, so that the counting of the number of sleepers is simpler. Then, when sleepers are respectively recognized at the connection of two adjacent sleeper images, it means that the same sleeper is split into two adjacent sleeper images and is detected by the recognition algorithm. At this time, one needs to be subtracted from the accumulated quantity. The artificial intelligence detection and recognition algorithm can also recognize the width of the sleepers in the sleeper image to assist in determining whether the same sleeper is split into two adjacent sleeper images.

[0036] Based on the above technical solution, optionally, as Figure 2As shown, the system further includes a GPS positioning module for mileage positioning. Specifically, the mileage positioning method based on the number of sleepers and the mileage positioning method based on GPS can be used for full-course positioning separately, or can be combined for positioning specific partial distances. Optionally, the GPS positioning module adopts a positioning method that combines GPS and high-precision inertial navigation. Specifically, the extended Kalman filter algorithm can be used to correct the integral value of the IMU according to the GPS data.

[0037] Further optionally, the target driving period is the period during which the GPS positioning module has abnormal positioning. Specifically, global positioning can be performed through the GPS positioning module. The GPS positioning module is connected to the data analysis module 200 (such as Figure 2 the industrial computer in it). The data analysis module 200 can obtain the positioning signal of the GPS positioning module in real time. When the positioning signal strength of the GPS positioning module is weak, there are no satellites, or the HDOP horizontal precision factor is too large (such as greater than 20), it can be determined that the GPS positioning module has abnormal positioning, that is, it enters the target driving period. At this time, the number of sleepers can be counted to perform auxiliary positioning based on the number of sleepers and determine the mileage information during the abnormal positioning period of the GPS positioning module. Thus, stable long-distance auxiliary positioning can be achieved in the case of lost global positioning, making the positioning more accurate.

[0038] Based on the above technical solution, optionally, as Figure 2 shown, the system further includes a railway inspection module for detecting railway defects. Specifically, the railway inspection module can include various defect detection units, such as a track profile wear detection unit, a track smoothness detection unit, a rail surface damage detection unit, a rail internal damage detection unit, etc., and any existing detection method can be used to implement it. The railway inspection module is connected to the data analysis module 200 (such as Figure 2 the industrial computer in it). The data analysis module 200 can summarize the detection results of the current railway inspection module in real time.

[0039] Further optionally, the data analysis module 200 is further configured to receive the defect information sent by the railway inspection module, and record the sleeper images of each preset number collected before and after the corresponding moment together with the corresponding defect information. Specifically, when the railway inspection module detects a defect, it can generate defect information and send it to the data analysis module 200. When receiving the defect information, the data analysis module 200 can separately save the sleeper images of each preset number (such as 20) collected before and after the current moment in real time, and record them together with the corresponding defect information in an independent folder. In particular, for defects such as track profile wear, internal damage of the rail, and track smoothness, which cannot be judged by the naked eye, due to possible mileage positioning errors or errors of the positioning devices carried by the staff when going for maintenance, the staff may not be able to accurately locate the defects. By accurately recording the sleeper images near the defect location when the defect is detected, the auxiliary positioning of the defect is realized to guide the staff to accurately locate the defect location.

[0040] Based on the above technical solution, optionally, the data analysis module 200 is further configured to detect the buckle and sleeper defects according to the sleeper images. Specifically, the data analysis module 200 can use the buckle and sleeper defect detection algorithm for detection, and can specifically detect defects such as buckle displacement, buckle loss, and sleeper cracks. Among them, the neural network structure adopted by the buckle and sleeper defect detection algorithm is as Figure 3As shown, it successively includes two CBS modules, a CSP1 module, a CBS module, a CSP2 module, a CSP3 module, and a convolutional layer conv to obtain a prediction result. Among them, the CBS module successively includes a convolutional layer conv, a normalization layer BN, and an activation function layer SiLU. In the CSP1 module, on the one hand, the input passes through a CBS module, a residual unit res unit, and a convolutional layer conv in sequence and then is input into the concat fusion unit. On the other hand, it is input into the concat fusion unit through a CBS module. The output of the concat fusion unit passes through a CBS module and is used as the output of the CSP1 module. In the CSP2 module, the input passes through a CBS module, two residual units res unit, and a convolutional layer conv in sequence and then is input into the concat fusion unit. Among them, the output of the CBS module also passes through a CBS module and is input into the concat fusion unit. The output of the concat fusion unit passes through a CBS module and is used as the output of the CSP2 module. In the CSP3 module, the input passes through three CBS modules and a convolutional layer conv in sequence and then is input into the concat fusion unit. Among them, the output of the first CBS module also passes through a CBS module and is input into the concat fusion unit. The output of the concat fusion unit passes through a CBS module and is used as the output of the CSP3 module. The residual unit res unit includes two CBS modules and an add fusion unit connected in sequence. On the one hand, the input passes through two CBS modules and is input into the add fusion unit. On the other hand, it is directly input into the add fusion unit. The output of the add fusion unit is used as the output of the residual unit res unit.

[0041] Before using the above neural network, it can be trained first. Specifically, an input resolution of 640×480 can be used, and image data collected on-site can be used. Ten thousand training samples can be generated through data augmentation for training. The network can output 4 categories, including normal sleepers, cracked sleepers, normal fasteners, and misplaced fasteners. Further, it can be inferred whether there are missing fasteners, such as there need to be two fasteners in the unilateral rail sleeper area, etc.

[0042] The mileage positioning system of the railway inspection vehicle provided by the embodiment of the present invention includes a sleeper image capture module and a data analysis module. The sleeper image capture module captures a sleeper image every preset driving distance, and the data analysis module counts the total number of sleepers during the target driving period according to the sleeper image, and then determines the mileage information during the target driving period according to the total number of sleepers. More accurate mileage positioning can be achieved, and the problem of positioning loss is not likely to occur. Further, the functions of railway defect auxiliary positioning and buckle and sleeper defect detection can also be simultaneously realized through the sleeper image capture module and the data analysis module.

[0043] Note that the above is only a preferred embodiment of the present invention and the technical principles applied. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, re-adjustments and substitutions can be made by those skilled in the art without departing from the protection scope of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments. Without departing from the concept of the present invention, more other equivalent embodiments can be included, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A mileage positioning system for a railway inspection vehicle, characterized in that, The system includes: a sleeper image capture module and a data analysis module; wherein, the sleeper image capture module is used to capture a sleeper image every preset driving distance; the data analysis module is used to count the total number of sleepers during the target driving period according to the sleeper image, and determine the mileage information during the target driving period according to the total number of sleepers.

2. The mileage positioning system of the railway inspection vehicle according to claim 1, characterized in that, The sleeper image capture module includes an encoder, a microcontroller and a camera; wherein, the encoder is used to generate a pulse signal based on the rotation of the wheels of the inspection vehicle; the microcontroller is used to count the driving distance according to the pulse signal, and generate a shooting trigger signal every time the preset driving distance is counted; the camera is used to trigger the shooting of the sleeper image according to the shooting trigger signal and upload it to the data analysis module.

3. The mileage positioning system of the railway inspection vehicle according to claim 1, characterized in that, The data analysis module is specifically used for: determining the mileage information according to the product of the total number of sleepers and the standard sleeper spacing.

4. The mileage positioning system of the railway inspection vehicle according to claim 1, characterized in that, The two actual railway areas corresponding to two adjacent sleeper images are continuous; the data analysis module is specifically used for: using an artificial intelligence detection and recognition algorithm to real-time identify the number of sleepers in each sleeper image; adding up the number of sleepers, and if sleepers are respectively identified at the joints of two adjacent sleeper images, subtracting one from the statistical quantity to obtain the total number of sleepers.

5. The mileage positioning system of the railway inspection vehicle according to claim 1, characterized in that, The system further includes a GPS positioning module for mileage positioning.

6. The mileage positioning system of the railway inspection vehicle according to claim 5, wherein, The target driving period is the period during which the GPS positioning module has abnormal positioning.

7. The mileage positioning system of the railway inspection vehicle according to claim 5, characterized in that, The GPS positioning module adopts a positioning method that fuses GPS and high-precision inertial navigation.

8. The mileage positioning system of the railway inspection vehicle according to claim 1, wherein The system further includes a railway inspection module for detecting railway defects.

9. The mileage positioning system of the railway inspection vehicle according to claim 8, characterized in that, The data analysis module is further used to receive the defect information sent by the railway inspection module, and record the corresponding defect information together with each preset number of sleeper images collected before and after the corresponding moment.

10. The mileage positioning system of the railway inspection vehicle according to claim 1, characterized in that The data analysis module is further used to detect the buckle and sleeper defects according to the sleeper image.

Images